High-throughput material search by magnetic and compositional mapping of reactively sputtered combinatorial Fe_xV_yN_z films

Combinatorial sputtering is a deposition technique that exploits unique cosputtering geometries to obtain analog compositional spaces across entire wafers.^1–3 This family of techniques has been widely leveraged to investigate novel high entropy alloys,⁴ photovoltaics,⁵ piezoceramics,⁶ battery components,⁷ and magnetic² materials to optimize properties. While synthesis of combinatorial films can be straightforward, high-throughput characterization remains a bottleneck for both acquisition and analysis. For example, many commonly employed compositional measurement techniques, including energy-dispersive x-ray spectroscopy (EDS) and x-ray photoelectron spectroscopy (XPS), are challenged to quickly and reliably map a composition across an entire 6–8 in.-diameter wafer,^8,9 especially in the presence of dilute species or elements with low-Z, including B, N, and O.¹⁰ Accordingly, such investigations frequently leverage synchrotron resources to complete high-fidelity property mapping, which provide the required probe intensities and translation capabilities.^9,11

Among the compositionally complex materials spaces within which combinatorial synthesis presents throughput advantages are doped iron nitrides (Fe_xM_yN_z), which are currently under investigation for application as next-generation rare earth-free permanent magnets^12–14 as well as magnetic read–write heads,^15,16 spin injecting components in spintronics,¹⁷ and oxygen evolution catalysts.¹⁸ Generally, candidate materials for rare earth-free permanent magnet applications, including Alnico,¹⁹ MnAl,²⁰ and FeN (α″-Fe₁₆N₂),^13,21 among others,²² are those that display large and thermally robust maximum energy products [(BH)_max]. For example, experimentally realized (BH)_max values for components of Alnico, MnAl, and α″-Fe₁₆N₂ are 10.5,²² 7,²³ and 20²⁴ MGOe, respectively, at room temperature, which are competitive with but remain less than the 54²⁵ MGOe reported for NdFeB. Motivated primarily by the existence of this metastable, high (BH)_max α″-Fe₁₆N₂ phase,²⁶ significant research has been devoted to the development of FeN and its doped counterparts, specifically to understand the role of chemistry on the thermal stability,¹⁶ magnetic properties,²⁷ and nitrogen diffusivity^15,28 of these materials. For example, a comparison of Al and Zr dopants (3–5 at. %) on lightly nitrided Fe showed that inclusions reduced Fe and N self-diffusion and accordingly improved the temperature stability of each alloy.²⁸ In general, such investigations rely on preparation of individual samples with digital compositions and subsequent magnetic characterization^14,27–30 to determine the effects of the dopants on the phase assemblage and magnetic properties. In particular, the inclusion of V has been shown, using density functional theory, to potentially affect the relative free energies of the low-N FeN phases.^17,31 Further, incorporation of V into lightly nitrided FeN films has been experimentally shown to drive changes in lattice parameters,¹⁴ although the effects of such doping on the magnetic properties remain to be investigated. While experimental investigations of other dopants, including Co,^32,33 Mn,³⁰ Al, Zr,²⁸ and Ti²⁹ have revealed positive effects on thermal stability and saturation magnetization (magnetization at maximum applied field, M_s) of the α″ phase, such reports for V-doping remain outstanding. Moreover, given the number of potential dopants (and co-dopants) that have shown to be impactful on such characteristics,^28,31 these and similar compositionally complex doped magnetic systems stand to benefit from the application of high-throughput combinatorial synthesis and characterization procedures.

In spite of suitable detection limits for measuring lightly nitrided Fe_xM_yN_z films,¹⁰ the long sample/environment preparation, acquisition, and analysis times make conventional laboratory-based XPS challenged to effectively map the composition of wafer-scale areas. Additionally, XPS is extremely surface sensitive and often requires milling procedures to remove adventitious and native carbon and oxide layers if the measurement is completed ex situ.^34,35 Such milling procedures take time for each spatial point acquisition, and may require optimization for different materials.³⁶ While EDS is not as surface sensitive as XPS (i.e., it can be optimized for surface or bulk sensitivity) and does not generally require any milling procedure,³⁷ such measurements are usually confined to the horizontal field width of the scanning electron microscope (SEM) magnification and it can be more challenged to quantify dilute, lighter atomic species. For example, modern silicon drift detectors (SDDs) employed in EDS systems report errors of ∼2–5 at. % in dilute (i.e., 25–33 at. % nitrogen) nitrides when analysis is conducted without standards, while error is reduced when compounds are more heavily nitrided.³⁸ In contrast to XPS and EDS, x-ray fluorescence (XRF) boasts the spot size and translatability, bulk sensitivity, and acquisition times to meet the high-throughput requirements of combinatorial investigations.³⁹ Moreover, the recent application of wavelength dispersive systems [Wavelength Dispersive X-ray Fluorescence (WDXRF)]^40,41 has circumvented challenging peak overlaps, which had traditionally made quantification of specific species pairs using non-dispersive XRF difficult.³⁹ Despite displaying the requisite sensitivities to dilute and light species and throughput characteristics to address the compositional measurement bottleneck for combinatorial synthesis investigations, WDXRF remains under-applied to wafer-scale compositional mapping.

In parallel with composition, characterization of magnetic properties at the wafer-scale can be experimentally challenging. Previous combinatorial investigations of magnetic materials, such as Fe_xAl_1−xN_yO_1−y, have relied on targeted superconducting quantum interference device (SQUID) measurements of specific small-area regions extracted from wafers to directly examine effects of composition on characteristics such as moment, coercivity, and remanence.⁴² Similar information has also been obtained through vibrating sample magnetometry (VSM) measurements of individual coupons from combinatorial magnetic multilayer FeCoB films.⁴³ Despite providing high-fidelity magnetic data, SQUID and VSM measurements are time consuming and are challenged to use to meet the throughput needs of combinatorial synthesis.^43,44 In contrast to SQUID and VSM, use of precision, translatable Magneto-Optical Kerr Effect (MOKE) boasts extremely high throughput⁴⁵ and has been recently employed to investigate the magnetic properties of combinatorially sputtered thin films.^9,44 This technique examines magnetic properties using the rotation of linear polarized light due to the presence of surface magnetization, and readily meets throughput the needs of combinatorial synthesis. As such, its application to doped magnetic thin films with accompanying WDXRF mapping provides a characterization route capable of quickly identifying compositions of interest for further study and development.

In this study, a high-throughput reactive combinatorial sputtering and characterization procedure is leveraged to investigate the effect of V inclusion on the magnetic behavior of lightly nitrided FeN. This procedure employs WDXRF to map film composition as a function of location, which is then correlated with magnetic properties, including room temperature coercivity and saturation magnetization using MOKE measurements made at each position across the wafer. Through this mapping, regions of interest are identified for further, targeted magnetic, chemical, and structural characterization using VSM, XPS, and x-ray diffraction (XRD). The magnetic and compositional properties observed through mapping are shown to be in close agreement with more targeted characterization techniques, supporting the capability of WDXRF and MOKE mapping for high-throughput applications such as combinatorial sputtering. Through these methods, it is revealed that incorporation of V reduces the saturation magnetization and may increase the coercivity of lightly nitrided FeN with moderate levels of inclusion (5–10 at. %), likely contributed to by a combination of vanadium nitride (VN) secondary phase formation and processing-imparted crystallographic texture. At inclusion levels beyond 18 at. % V, the macroscopic room temperature ferromagnetic properties of FeVN are observed to be completely reduced. Further, a high nitridation potential of V compared to Fe plays a strong role in nitrogen distribution throughout the wafer.

Combinatorial synthesis of Fe_xV_yN_z was conducted using a modified AJA ATC Orion sputter deposition system equipped with defocused AJA A315-LP guns with a base pressure of 4 × 10⁻⁸ Torr. Reactive DC deposition was conducted by simultaneously providing 1.14 and 0.96 W cm⁻² of power across 3.81 cm-diameter pure Fe (99.95%) and V (99.95%) targets (ACI alloys, Inc.), respectively, using a Pinnacle DC power supply (Fe) and a DCXS-750-4 Multiple Sputter Source DC power supply (V). For this deposition, the 6 in. (001) Silicon substrate was lowered to a position ∼6 cm above the guns. While combinatorial film thicknesses are variable across substrate surfaces, deposition powers were calibrated to achieve a thickness of ∼300 nm near the center of the wafer. The deposition was carried out at ambient temperature for 200 min in an atmosphere that consisted of 75 sccm Ar and 2 sccm of N₂, fed through Trigon Technologies indicating oxygen trap and regulated down to 4 mTorr using a Vakuumventile AG (VAT) gate valve paired with a VAT adaptive pressure controller. As a control, a deposition of FeN was conducted with identical conditions without turning on the V power source.

Following deposition, film compositions were measured across both combinatorial wafers using a Rigaku ZSX Primus 400 WDXRF equipped with LiF, RX, and PETH analyzing crystals and a Rh x-ray source. For Fe and V intensities, a LiF (200) crystal coupled to a proportional detector was utilized, whereas for N and O, RX45 and RX26 synthetic multilayer crystals were employed, respectively, both coupled to a P10 gas flow proportional count detector. An aperture was used to regulate the incident spot to a translatable 10 mm region, which was incremented at distances of 7.5 mm across the wafer (2.5 mm overlap) for a total of 277 points. Each wafer took ∼24 h to measure an entire map (four species total), with the integration of each region for ∼3 and ∼2 min of translation overhead. A thin film model was assumed for this analysis, and standards were not used for quantification. For a comparison with WDXRF composition values, regions of interest on both combinatorial films were additionally measured using a Thermofisher Nexsa XPS equipped with a Al Kα source. For these measurements, a 400 μm diameter spot size was employed to collect the photoelectron yield from the N 1s, O 1s, V 2p, and Fe 2p binding energy regions following a 120-s-long Ar milling procedure (4000 eV, “high current” mode) to remove native surface oxide. The associated data were fit and analyzed using the Thermo Avantage software suite to determine atomic compositions. Film compositions within regions of interest were also characterized using an Oxford X Max^N energy-dispersive x-ray spectroscopy (EDS) system in a Zeiss Gemini 560 SEM and fit using the automated routines in the associated Oxford Aztec software. Each measurement was conducted by integrating a 90 μm horizontal field width (1.3 k× magnification) image for 20 min using a 15 kV accelerating voltage.

MOKE mapping was completed using a Durham Magneto Optics Ltd., NanoMOKE 3 setup. A 660 nm laser diode was focused to ∼10 μm at a 45° angle to the surface normal of the 6 in. wafer. The intensity and circular rotation of the reflected beam were recorded simultaneously while sweeping a magnetic field to ±300 Oe at 1 Hz. The wafer was mounted to two crossed Thor Labs Linear Translation Stage (LTS) 150 mm tracks and mapped with 2 mm steps. In addition, in-plane, room temperature VSM measurements were made up to 4 kOe on each ∼5  × 5 mm² region of interest using a Lakeshore system equipped with a 648 Electromagnet Source. Film volumes were calculated by measuring the exact film area through the use of an optical camera image and ImageJ analysis software and multiplying by the film thickness, measured with SEM (Zeiss Gemini 560) images of the edges.

XRD measurements were made on each region of interest using a Bruker D8 Discover X-ray Diffraction System with a Dectris Eiger 2R 500 K area detector and rotating anode Cu Kα source. Diffracted intensity between 10° and 100° in 2θ was collected in a parallel beam symmetric geometry using a Goebel mirror with an affixed 0.5 mm collimator. The area detector, which has a ψ-width of ±3.4°, was summed in the $ψ$ direction to produce conventional $2θ–ω$ diffraction patterns. Peaks observed within the integrated patterns were subsequently fit to Pearson VII peak shapes using Line-Profile Analysis Software diffraction analysis software⁴⁶ to determine their positions. Pole figures were collected on the (110) FeVN peaks within select regions of interest. For these measurements, the detector was left to integrate for 30 s at each configuration while $ψ$ incremented at angles spanning between 0° and 75° and $ϕ$ between 0° and 360°. Pole figure processing was completed using Bruker DIFFRAC Texture analysis software. Further, each region of interest was subjected to atomic force microscopy (AFM) measurements with 1  × 1 μm² field widths made using a Bruker Asylum system in tapping mode. AFM images were processed using Gwyddion analysis software to determine relative surface roughnesses.

The reactive combinatorial sputter deposition process employed to grow Fe_1−xV_xN_y films is diagramed in Fig. 1. Compared to a typical reactive DC sputter deposition process, several modifications have been made to facilitate a large compositional spread across the 6 in. Si wafer. For example, the angle of incidence of each gun is set to 9.5°, producing a deposition focal point at a height of 39 cm. Further, the wafer height is set at 9 cm, well below the focal point, so that the deposition focal of each individual gun is trained ∼5 cm away from the wafer center. The employed targets have a small diameter (3.84 cm) to further accentuate the distance between the growth centers of each species. Finally, to achieve a combinatorial process, the wafer is left stagnant (no rotation). These characteristics enable the preparation of combinatorial depositions with maximized compositional spread. Moreover, the Ar-rich deposition atmosphere (Ar:N₂ flow = 75:2) and powers of 1.14 and 0.96 W cm⁻² used for Fe and V, respectively, are optimized to achieve a low-incident ion energy growth of FeN underneath the Fe sputter gun with a dilute V-doping in an immediate vicinity. For the FeN control wafer, the V source remained off for the duration of the deposition.

Following deposition, the Fe, V, and N atomic compositions were mapped using wavelength dispersive x-ray fluorescence (WDXRF) across the entire 6 in. wafer, as shown in Figs. 2(a)–2(c). The WDXRF map of the FeN wafer is provided in Fig. S1 in the supplementary material. This technique collects the characteristic photons produced within the film due to an incident x-ray beam and is capable of measuring a 10 mm-diameter, translatable spot to characterize elemental composition as a function of position. Further, diffracting crystals (LiF, RX45, and PETH, depending on the fluorescence photon wavelength and expected fluorescing elements) are leveraged to circumvent peak overlaps and increase atomic resolution, allowing for unambiguous elemental identification and quantification. It should be noted that this technique quantifies the raw weight of each characterized species on a given area of surface, which is then converted to atomic composition using atomic weight.

Due to the combinatorial deposition conditions described previously, the composition of the film varies, between the Fe (denoted as stars in Fig. 2) and V (denoted as pentagons in Fig. 2) focal points, from Fe_84.1V_4.8N_11.1 to Fe_14.8V_37.9N_47.2, as plotted in Fig. 2(d). In the vicinity of the Fe focal point, WDXRF measures a N composition around 11 at. % (the requisite content for the $α ′$ and $α ′ ′$ phases) with a V-doping level of ∼5 at. %, indicating that this process has achieved the targeted compositions. Further, the film displays a N composition of ∼47 at. % near the V focal point, which is evidence that a highly nitrided V_xN_y phase has formed, which may display a nitridation potential larger than FeN. An additional region of interest is annotated on the WDXRF maps with a triangle, which will be later shown to display noteworthy MOKE properties. A final region of interest is identified through annotation with a diamond symbol on the WDXRF map of the FeN combinatorial wafer shown in Fig. S1 in the supplementary material. This additional region displays a WDXRF composition of Fe_91.6N_8.4 and serves as a comparison with the Fe_xV_yN_z regions of interest from the FeVN combinatorial wafer.

To correlate magnetic properties with the compositional map provided by WDXRF, MOKE loops were measured with a 10 μm diameter probe across the 6 in. combinatorial wafer in 2 mm steps. The Kerr rotation of the optical probe with application of a +300 Oe field as a function of film position is shown in Fig. 3(a), where the map has been rotated counterclockwise to position the Fe focal point at the bottom of the panel. Comparing this map to the compositional maps shown in Fig. 2, it is evident that the Kerr amplitude (proportional to saturation magnetization) strongly correlates with Fe composition, which is expected because Fe is the only ferromagnetic element present in the film. The same behavior is observed for film coercivity, mapped in Fig. 3(b), which also correlates strongly with the presence of Fe. For coercivity measurements, the applied field scales have been normalized from maxima at +300 Oe. In Fig. 3(c), the measured Kerr amplitudes and coercive fields of the film between the Fe (star point) and V (pentagon point) focal points are plotted as a function of film composition, which has been logged using the WDXRF map shown in Fig. 2(b). Both of these properties decrease to zero as the V composition approaches 18%, where the Kerr amplitudes display a gradual fall in contrast to the sharp reduction in coercive field. These observations indicate that inclusion of V in excess of ∼18% in lightly nitrided Fe films results in a deleterious vanishing of macroscopic ferromagnetic properties.

To the left side of the Fe-rich region in Fig. 2(b), an area with a high MOKE coercivity is observed, which boasts the highest values observed across the entire wafer. This region is marked with a triangle in Figs. 3(a), 3(b), and 2(a)–2(c), displaying a composition of Fe_74.3V_10.3N_15.4. While the maximum Kerr amplitude of this region is low [Fig. 2(a)], indicative of a low saturated magnetic moment, it boasts a slightly higher V composition than the film present directly under the Fe focal point (annotated with a star). A comparison between MOKE loops measured at the most Fe-rich point [labeled with a star in Figs. 3(a), 3(b), and 2(a)–2(c)] and the high coercivity region with slightly higher V-doping [labeled with a triangle in Figs. 3(a), 3(b), and 2(a)–2(c)] is shown in Fig. 3(d). The loop measured on the Fe_74.3V_10.3N_15.4 composition displays a greater coercivity and lower maximum Kerr rotation than the same measurement on the Fe_84.1V_4.8N_11.1 composition, both of which represent regions of interest for further study.

Further compositional and structural analyses of the identified regions of interest [star, triangle, and pentagon points in Figs. 2(a)–2(c), 3(a), and 3(b) and diamond point in Fig. S1 in the supplementary material (Fe focal point on FeN wafer)] were undertaken to explore the influence of dilute V-doping on the magnetic properties of FeVN and allow for comparison between high-throughput mapping and more targeted characterization techniques. To accomplish this characterization, ∼1 cm² coupons at the positions identified as regions of interest through WDXRF and MOKE mapping were manually cleaved out of the parent 6 in. wafers and submitted to further vibrating sample magnetometry (VSM), x-ray diffraction (XRD), x-ray photoelectron spectroscopy (XPS), and energy-dispersive x-ray spectroscopy (EDS) characterization.

To quantify the exact saturation magnetizations and coercivities of regions of interest, VSM measurements were carried out on each, as shown in Fig. 3(e). As was observed through MOKE response mapping, the M_s of each region of interest decreases as the amount of V increases. As such, the Fe_91.6N_8.4 coupon displays the highest M_s of 1206 emu cm⁻³, followed by the Fe_84.1V_4.8N_11.1 coupon, within which a M_s of 1138 emu cm⁻³ was measured, with the Fe_74.3V_10.3N_15.4 and Fe_14.8V_37.9N_47.2 boasting further decrease of M_s of 714 and 17 emu cm⁻³, respectively. Film coercivity was found to decrease from 125 to 97 Oe as the composition changed from Fe_91.6N_8.4 to Fe_84.1V_4.8N_11.1. However, incorporation of additional V, up to a composition of Fe_74.3V_10.3N_15.4, was found to increase the film H_c to 165 Oe, representing a change of +70% concomitant with the additional 5.5 at. % V. This change compares similarly to the ∼50% increase observed using MOKE response mapping, which is shown in Fig. 3(d) for these regions of interest. Changes in coercivity observed using both MOKE and magnetometry may be driven by several factors, including the formation of a VN secondary phase,^47,48 a change in exchange interaction between Fe species due to the presence of substituted V, or differences in crystallographic texture. A further assessment of the stoichiometry, chemical environment, and corresponding phases, of each region of interest, as well as a comparison between mapping and targeted compositional measurements, was provided by EDS and XPS measurements.

XPS spectra were collected in the Fe 2p, V 2p, O 1s, and N 1s binding energy regions on each coupon following 120 s of Ar milling into the film center. Example spectra collected from the star region of interest (displaying a WDXRF composition of Fe_84.1V_4.8N_11.1), with their associated fits are shown in Figs. 4(a)–4(c). Analogous XPS spectra and analyses for the other regions of interest are shown in Fig. S2 in the supplementary material. A comparison between high-throughput WDXRF and targeted EDS and XPS composition measurements is provided in Table I. Within each region of interest, where WDXRF and XPS report similar atomic compositions to within <5 at. %, irrespective of the level of dilution or atomic weight of the quantified species, EDS-derived stoichiometries are more varied. For example, qualitatively, WDXRF estimates a ∼3 at. % higher N composition and a ∼3 at. % lower Fe composition than XPS, while both techniques report essentially identical V compositions. By comparison, EDS measurements of the regions of interest within the combinatorial wafer recover nitrogen compositions that are 5–20 at. % higher than both other techniques. Accordingly, EDS-measured Fe and V compositions of these regions also comparatively vary by 2–20 at. %. For these calculations, oxygen was omitted as a constituent element. While it was detected in EDS and readily observed in XPS measurements for each sample at every depth of Ar milling (as detailed in Fig. S3 in the supplementary material for the star region of interest), the XPS signal was observed to consistently decrease with increased milling time. Given the surface sensitivity of the technique and decreasing O 1s intensity with Ar etch time, this oxygen signal is likely an artifact from the native surface oxide being implanted deeper into the film during the milling. EDS detection of oxygen in each region of interest also likely originates from the native oxide, where the similar energy between the O Kα and V Lα characteristic x rays may further affect atomic composition calculations. Moreover, WDXRF did not detect a quantifiable amount of oxygen, which is expected because the interaction volume of the probe is much larger than the thickness of the film, and not specifically sensitive to the surface. Based upon this comparison, it is evident that WDXRF represents a suitable high-throughput characterization method to map the compositions of combinatorially deposited wafers with dilute, light species that may have ambient oxide layers.

Aside from composition quantification, XPS spectra provide a means to analyze the chemical environment of each species through the examination of shifts in binding energy and changes in amplitude. The locations of each peak measured for each region of interest are provided in Table SI in the supplementary material. Fe 2p peaks were found at binding energies indicative of both α-Fe and Fe_xN bonding, which arise in identical locations,⁴⁹ whereas V peaks were found in locations consistent with V–N bonding.^50,51 Peaks from all three species were observed to shift to slightly higher binding energies (+0.2 eV) with increased V composition, which may be related to slight changes in local electronegativity.⁵² Further, the small magnitude of this shift is indicative of negligible charge transfer between the species or changes in ionization.⁵³ Each N 1s region was found to contain two peaks each in positions that are typical of nitrogen atoms bonded into nitrides. In each case, the stronger peak is located at ∼397.5 eV, which has been experimentally shown to correspond to α′/α″-FeN,⁵⁴ γ-Fe₄N,^49,55 and VN,^50,56 making unambiguous determination of the bonding for nitrogen difficult in these spectra. However, this peak was observed in the N 1s spectrum measured from the Fe_91.6N_8.4 region of interest [Fig. S2(c) in the supplementary material], indicating that observed intensity at this binding energy in all samples may be, at least partially, attributed to Fe–N bonding. As such, the presence of a secondary VN phase forming within the star and triangle regions of interest cannot be completely ruled out through XPS analysis. The lower intensity, higher binding energy peak in the N 1s regions has been associated with surface adventitious layers⁵⁰ and was found to decrease in intensity relative to the nitride peak as Ar milling proceeded (Fig. S3 in the supplementary material), indicating that it is related to the presence of the adventitious oxide. In total, analyses of XPS spectra reveal Fe 2p peak locations consistent with both α-Fe and/or FeN, V 2p peak locations indicative of VN formation, and N 1s locations consistent with both FeN and VN bonding, as well as artifacts from surface oxide removal. While it is not possible to delineate whether nitridation of Fe or V has dominated in the Fe-rich regions of the FeVN wafer (i.e., if a VN secondary phase is responsible for the nitrogen bonding), the N 1s peak is observed at an identical location in the FeN wafer.

Further structural analysis of the combinatorial film is provided by examination of integrated area detector $2θ–ω$ diffraction patterns measured on coupons from each region of interest, as shown in Fig. 5. Each pattern contains a strong peak corresponding to the Si substrate at ∼69° (Cu Kβ) and displays broad peak widths suggestive of nanocrystalline grains. The patterns collected on the star (Fe_84.1V_4.8N_11.1, directly under the Fe focal point) and triangle (Fe_74.3V_10.3N_15.4, high MOKE coercivity) regions of interest both display peaks consistent with a primary α-Fe $Fm 3 ¯m$ phase, as does the Fe_91.6N_8.4 coupon from the FeN wafer. However, the peaks present in the patterns collected on the Fe_84.1V_4.8N_11.1 and Fe_74.3V_10.3N_15.4 coupons display shifts to lower angles that increase in magnitude with additional V composition, which is indicative that the incorporation of V has expanded the unit cell or strained the film. The change in each peak location with V composition, along with the corresponding change in c/a ratio and lattice volume, is shown in Fig. S4 in the supplementary material. A negligible change in c/a is calculated using each pattern, whereas the lattice volume increases with larger V inclusion. While increasing V substitution would be expected to drive such an expansion, given the larger lattice volume of metallic V compared to Fe, it is not possible to directly decouple the effects of peak shifting from strain in this analysis. However, the shifts in peak positions are consistent with as much as $ε≈ 2 %$ with 10.3 at. % V compared to the ideal Fe structure, which is greater than would be expected due to microstructural strain alone. Therefore, the volumetric expansion observed using XRD supports that the incorporated V is at least partially substituting for Fe. However, the presence of anisotropic strains, which would be expected in a textured thin film (as will be discussed later), may also contribute to peak shifts. The diffraction pattern collected on the region of interest cleaved at the pentagon point, displaying a WDXRF composition of Fe_14.8V_37.9N_47.2, contains three film peaks consistent with a $Fm 3 ¯m$ VN phase.⁵⁰ None of the patterns display any peaks consistent with the presence of α′-Fe₈N, $α ′ ′ - F e 16 N 2$⁠, γ′-Fe₄N, or ɛ-Fe_xN, which is indicative of either undetectably low quantities or complete lack of presence of these phases. A lack of α′-Fe₈N or $α ′ ′ - F e 16 N 2$ is expected due to the stringent deposition requirements for these phases, where the combinatorial process employed for this investigation does not leverage epitaxial lattice templating or facing target plasma geometries. Peaks corresponding to VN were not observed in the patterns measured on the star and triangle regions of interest, possibly due to small interaction volumes or grain sizes or low crystallinity of this potential secondary phase.

While it is clear that the increased coercivity at the triangle region of interest is detected through MOKE mapping and confirmed through VSM measurement, the reason for this property difference is not evident based upon these measurements. As the wafer is kept stagnant during deposition, each region of interest experiences a slightly different separation distance and angle with the Fe and V sputtering guns, which may affect crystallographic texture and morphology. Accordingly, AFM and pole figure measurements were carried out to compare the degree to which film texture and morphology vary across the wafer. Analysis of AFM images of each region of interest, provided in Fig. S5 in the supplementary material, reveals surface morphologies that do not qualitatively vary depending on location (and composition). Further, relative surface roughnesses do not strongly vary, all falling between 1.7% and 2.7%. Based upon this comparison, differences in film microstructure are likely not responsible for the increase in coercivity perpendicular to the composition axis. Pole figure measurements of the (110) reflections collected on the star and triangle regions of interest, however, do show unique textures that likely arise due to the deposition conditions, as shown in Figs. 5(b) and 5(c). Each pole figure displays a low diffracted intensity for the (110) peak near the film-normal direction, which is similar to what would be expected for a fiber-like texture. However, the position of this low intensity center is different between the films. Moreover, the star region displays an intensity ring around 30° in $ψ$⁠, whereas the triangle region does not have a distinguishing texture in this location and instead shows broken intensity spots at ∼90° intervals at $ψ$ between 45° and 60°. As expected, the different deposition conditions have produced a non-uniform crystallographic texture across the wafer, which likely contributes to the variation in coercivity perpendicular to the compositional gradient.

A high-throughput combinatorial synthesis workflow has been utilized to investigate the effects of doping on reactive direct-current sputtered, rare earth-free FeVN thin films. This procedure leverages WDXRF and MOKE response mapping to quickly correlate differences in chemistry with magnetic behavior. The observations made using these high-throughput measurements have been compared with more targeted compositional and magnetic characterization of regions of interest using XPS and VSM, reporting nearly identical quantitative information. Through these methods, incorporation of V in FeN thin films has been found to decrease M_s, which vanishes with incorporation in excess of 18 at. %. Further, film coercivity was found to decrease with the inclusion of V up to ∼5 at. % and then increase in compositions up to ∼10 at. % V, likely contributed to by differences in processing-imparted crystallographic texture, followed by complete cessation along with saturation magnetization with further addition. While the formation of a VN secondary phase may possibly drive the observed changes in coercivity with increasing V composition, analysis of targeted XPS and diffraction measurements could not unambiguously identify their presence. Further, lattice volume calculations from diffraction peak positions supported that V substitution occurred within the Fe combinatorial film. This high-throughput process is enabling for the investigation of compositionally varied and complex magnetic alloys and is capable of quantitatively determining the chemistries of materials with dilute and light species. Further experimentation using this procedure may include the investigation of different dopants or other sputter-compatible magnetic alloy systems, as well as the use of temperature-resolved MOKE to additionally enhance the screening and exploration capability.

Additional information is available in the supplementary material, including WDXRF compositional mapping of the FeN control wafer, XPS composition measurement of additional FeVN and FeN combinatorial regions of interest, XPS intensity ratios for O 1s and N 1s peaks from adventitious oxygen and nitrogen species, plotted d-spacing shifts, c/a ratio and unit cell volume changes as a function of V-doping level, AFM measurements of each region of interest, and tabulated binding energy positions of XPS peaks for each region of interest.

This work was supported by resources from Niron Magnetics, Inc. in Minneapolis, Minnesota, USA.

A.M. and F.J. are researchers at Niron Magnetics Inc., a company commercializing Fe₁₆N₂.

Shelby S. Fields: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review & editing (equal). Olaf M. J. van ‘t Erve: Formal analysis (equal); Investigation (equal); Methodology (equal). Andrew McGrath: Conceptualization (equal); Writing – review & editing (equal). Francis Johnson: Conceptualization (equal); Writing – review & editing (equal). Steven P. Bennett: Conceptualization (equal); Funding acquisition (lead); Methodology (lead); Project administration (lead); Resources (lead); Supervision (equal); Writing – review & editing (supporting).

The data that support the findings of this study are available from the corresponding author upon reasonable request.